Analog Dialogue: Editor's Notes and Versatile New Module Y(Z/X)^m at Low Cost
Editor’s Notes
NEW PRODUCTS
We've never felt the need to
apologize for anything that has
appeared in Dialogue, and we
intend to keep matters that way.
However, we do think that fans
of our “Applications” section
deserve an explanation for its
truncation in favor of a veri-
table “raft” of New Productsin ===
Simply stated, our innovative “boys in the back room” have
gone-and-done-it again. There is something about the Spring
season (in sway at this writing) that quickens the processes of
life: after conception last year (or earlier) and a winter of ges-
tation, the budding and flowering occur with a suddenness and
massiveness that is breathtaking. And our creative colleagues,
similarly, seem to be graced with this Call of Nature.
And so, we suddenly have an ermbarrassingly-rich assortment of
new products, each of which seems —in its own way— to be the
ideal answer to some unheralded need. The significance of some
should be obvious, for example, the tremendous potential in-
herent in the utility and low cost of monolithic 10-bit DAC’s*.
Many of the others can be “sold” on the basis of salient im-
provements in specifications, packaging, or price, all of which
add up to increased value for the knowledgeable user.
Bur then there are some that have potentially-great significance
that is not immediately obvious because they ar: unfamiliar:
the type 433 ¥(/Z/X)™ device, the AD2002 2¥%digit (0.5%)
infra-low-cost DPM, the MDA-11MF fast multiplying DAC;
their novelty poses a challenge to the imagination. For this
reason, the stories describing them include suggestions for
applications that amount to gem-like little “Application Briefs."
And that is where much of this issue’s “Applications” section
is, not in one neat package, but widely dispersed ia the context
of the products that make the applications possible. Pechaps
these ideas, and products, will help you now plant the seeds
of new uses that may come to fruition by next Spring) ppp
TEASER UNMASKED
In Volume 5, No. 5, we offered this formidable expression and
suggested that it had a neat closed-form solution, that could
be embodied with a new multi-purpose module, soon to
be announced.
y |
W = f(U,V)=U+ Y (1)
2U+ v
2
2u+—Vv
QI vias
The closed-form solution (to dispense with suspense) is
w=JU7+V (2)
and it is achieved with low cost and good accuracy by the
Model 433 (described in the adjacent pages) and two op
"Watch for the AD560, to be announced in the next issue of Dialogue,
amps. If you solved the problem in the following way, you
were inevitably led to the scheme described below:
Recognize that 2U (indicated by the arrow) = U + U, and that
consequently
y?
We Ot OW @)
This expression is an implicit solution for W, containing the
key to the use of the 433, which we'll come back to in a
moment. The reat of the solution goes:
(W - U) (W+U)=V? (4)
whence
WwW? =U? +V? (5)
and w=/u+v?
Returning to equation (3), it is evident that if you have a device
that computes Y*Z/X, and two summing devices (op amps), a
configuration that embodies equation (3) can be drawn.
-W-Ue wWeu
W+u du
Y ¥2
V x
z
|
Ss 35
Then, if the input-signal polarities and scale factors are correct,
in terms of the devices used and the sign of the square-root, the
computation will be performed to good accuracy over a wide
dynamic range of cither (or both) input(s).
The conventional (and hardly satisfactory) configuration for
performing vector summation is simply to use two multipliers
(or squarers), followed by a square-root-connected multiplier/
problems. Configurations using squarers and rooters employing
logarithmic techniques salve the dynamic range problem, but
still have undesirable cost and complexity. The Model 433
would appear to be a better way.
A number of readers sent us correct solutions in terms of the
final equation, but they all used a method of substitution that
involved a (more-difficult) quadratic equation that completely
obscured the point established by equation (3), i.c., an implicit
aclution was possible, given the availability of a YZ/X device.
Dan Sheingold >>>
analog dialogue
Route 1 Industrial Park, P.O. Box 280, Norwood, Mass, 02062
Published by Analog Devices, Inc., and available at no charge to engi-
necrs and scientists who use or think abcut 1.C. or diserctc analog, con-
version, data handling and display circuits, Correspondence is welcome
and should be addressed to Editor, Analog Dialogue, P.O. Box 280,
Norwood, Massachusetts, U.S.A. 02062. Analog Devices, Inc., has
representatives and sales offices throughout the world, For informa
tian regarding our products and their applications, you are invited to
tse the enclosed Business Reply card, write to the above address, or
phone 617-329-4700, TWX 710-394-6377, Telex 710-394-6577, or
cable ANALOG NORWOOD MASS,
VERSATILE NEW MODULE: Y(Z/X)]™ AT LOW COST
LOG-ANTILOG TECHNIQUES ARE USED TO MULTIPLY, DIVIDE, EXPONENTIATE
0.5%(FS) ERROR AS A DIVIDER WITH 100:1 RANGE OF DENOMINATOR (NO-TRIM)
0.5%(FS) ERROR AS A SQUARE-ROOTER WITH 10,000:1 INPUT RANGE (NO-TRIM)
by Fred Pouliot and Lew Cours
In Volume 5, Number 5 of Dialogue, we predicted the intro-
duction of a new “LAMDE" module, LAMDE is an acronym
that implies a set of capabilities: Log, Antilog, Multiplication,
Division, Exponentials. However, the new Model 433" has
applicabiliry that goes so far beyond even this impressive
list, that we decided simply to call it by the unglamorous
but apt designation “Programmable Multi-Function Module.”
WHAT IT DOES
The equation describing its overal] functioning, with appropri-
ate external connections, is (for its 3 inputs, X, Y, Z)
10 V2 Vx, Vy, Vz, Eo > 0
Eo=-5 Vy (+ y"
Vx Sims
from which one can deduce that it will multiply, divide, raise
(or lower) ratios to an arbitrary power. In addition, it can be
connected to perform squaring, rooting, and —with a small
amount of external circuitry— rms and vector computation,
Terminals that are available for manipulation of the exponent
m also permit logarithmic outputs to be developed. To provide
scale constants for operations involving only pairs of variables,
or for operations where a low-drift reference voltage is (in
general) desirakle, a nominal 9V low-TC reference output
is available.
As a multiplier or a divider, maximum error, without trimming,
is in the 0.5% jof full-scale output) class, for a wid> range of
inputs. Typical error is 0.3% of output plus SmV. (See tabula-
tion on page 5.) Small-signal and large-signal bandwidth tend
to be proportional to input voltage, ranging from 50 or
100kHz downward.
WHERE TO USE IT
Typical application areas for the Model 433 are those that in-
volve manipulation of analog voltages in analog data-reduction
and computing, pre-processing before digitizing in data acquisi-
tion, and design of sophisticated measuring instruments,
Examples include “true-rms" measurements, computing vector
sums and differences, and —using the adjustable exponent—
ideal gas computations, curve fitting, linearizing, and developing
approximations for analog computation,
Figure 1. Functional Block Diagram of Model 433
“For complete information on the Model 433, use the reply card Request
Gt.
Pethaps the most important area of application is a quite
prosaic (but hitherto neglected) one: performing division of
analog voltages over wide dynamic ranges. When conventional
analog multipliers are used in feedback configurations for divi-
sion, errors and noise increase in inverse proportion to the
denominator voltage. Even if such a circuit has been arduously
trimmed for less than 0.05%(FS) error in one quadrant with
10V denominator, its error will be comparable to the 0.5%
guaranteed no-trim maximum error of Model 433 at 1V, and
ten times worse at 0.1V. The log/antilog techniques used in
Model 433 inherently yield small, nearly-constant errors (with-
ou: tweaking) that are essentially independent of the denomi-
nator level and are well-behaved near zero. This advantage be-
comes especially significant in such applications as square
rooting, when the output is fed back as the denominator input,
continued on page 4
Solution tothe Teaser . . . 2
Versatile New Module; Y(Z/X)™ ‘at Low Cost : 3
Low-Cost 10- and 12-Bit
Ultralinear Converters . . . . sss 7
Three New Sample-Holds: Economy,
Speed, Precision . . « « et 7
A Really Low Cost DPM, the AD2002. er to 9
Better Displays with Fast New
Multiplying DAC . . . on:sn 10
Low-Drift Instrumentation AmebGee. ; rs
Ultra-Fast, Low-Cost FET-Input Op Amp Sr 11
1.C. Insight: Not by DriftAlone. . 2... = 12
ad FH aoa
WithADC-10Z .... ihe 13
Application Brief: Cis Bastrasnentation,
Amplifier Drift . . .« pad .8 14
Worth Raiding! Thess Recent Boake. era 15
Four Recent Ads: 10uV/mo. Op Amp, Log
Amps, Instrumentation Amplifiers,
OApAL, Dusk PER ei eee 6